Netherlands Graduate School of Linguistics LOT Summer School 2006 Issues in the biology and evolution of language

1
Netherlands Graduate School of Linguistics LOT
Summer School 2006Issues in the biology and
evolution of language

• Massimo Piattelli-Palmarini
• University of Arizona
• Session 5 (and last) (June 16)
• New perspectives in the biology of language

2
A central consideration

• So much has changed in biology
• And in linguistic theory
• That
• I suggest we re-examine the whole issue of the
  biology of language and of reconstructing
  language evolution
• In search of new windows of opportunity

3
Foreword Beyond textbook genetics

• Science advances by opening windows of
  opportunity (Nadel Piattelli-Palmarini 2002)
• Mendelian patterns of genetic inheritance that
  produce neatly identifiable pathological
  phenotypes have been one such window (ever
  since Archibald E. Garrods Inborn Errors of
  Metabolism - 1908)
• Single point mutations in DNA that invariably
  correspond to identifiable phenotypic variations
  (phenylketonuria, sickle-cell anemia etc.)
• Such cases of high genetic penetrance are the
  substance of textbook genetics (biochemical
  pathways, hemoglobin)
• But they are not the typical or average gene
• Today, we are in need of other windows
• To go beyond this wonderful, but limiting, scheme

4
Some paradigmatic historical windows

• The motion of celestial bodies
• The emission and absorption spectra of gaseous
  substances
• The spontaneous emanations of uranium
  composites
• The diffraction of X-rays by organic crystals
• The genetics of bacteria and their viruses
• Inborn errors of metabolism
• The giant squid axon (Hodgkin and Huxley)
• The receptive fields of neurons in the primary
  visual cortex (Hubel and Wiesel)
• Grammaticality judgments of native speakers
• Mirror neurons

5
Another central consideration

• Obviously, we want a functional account and an
  evolutionary reconstruction of mind and language
• That is completely mechanical
• Like the ones, say, for the adaptive immune
  system, or for the eye
• No less mechanical than these (all things
  considered)
• But also not more mechanical than these,
• Given the known remarkable dynamic subtleties
  involved.

6
The new genetics
7
A brief summary

• Master genes (homeoboxes etc.)
• The ubiquity of pleiotropic effects (ex. for Otx
  the cerebral cortex, kidneys, testes, gut lining)
• Gene duplication, gene conversion and
  transposable elements
• RNA editing and quality control (chaperonines)
• The importance of non-coding sequences
• RNA-only genes
• Retroposons and SINEs (David Haussler UCSC,
  conserved from the coelacanth to humans)

8
Matthew Ronshaugen, Nadine McGinnis William
McGinnis Nature 2002
the first experimental evidence that links
naturally selected alterations of a specific
protein sequence to a major morphological
transition in evolution.
High sequence conservation and major phenotypic
differences coexist.
9
A brief summary - continued

• Micro RNAs and Argonaute proteins (ubiquitous
  gene expression modulators)
• The histone code (epigenomics)
• DNA methylation
• Kissing chromosomes and gene-expression waves
• Epigenetic mechanisms
• Cut and run mRNA (nuclear speckles)
• Ancestral DNA caches (revertant variants that
  have come from one of the great-grandparents,
  even if the immediate parent did not contain the
  variant) (Lolle and Pruitt, 2005)
• Previously unsuspected degree of individual
  genetic variation

10
Where it all started
1953
11
Classical genetics

• DNA sequence-based genetics the information
  contained in DNA is all encoded in the sequence
• The central dogma of molecular biology DNA makes
  RNA makes protein
• (Less than 2 of the human genome is made of
  protein-coding DNA)

protein
DNA
12
A missing dimension Plasticity
fidelity of conversion
plasticity of response
specificity of information
13
The importance of plasticity

• differential patterns of gene expression.
• Many of these differences arise during
  development and are subsequently retained through
  cell division.
• Cell fate a single cell precursor can develop
  into different cell types.
• Cell memory mother cells can transmit to
  daughter cells a functional state acquired in
  response to an endogenous developmental program
  or exogenous stimuli.
• Modifications in cell function induced by the
  environment.

14
(No Transcript)
15
Main new lessons (so far)

• An organism does not inherit a naked DNA
• But an entire, organized and partially
  pre-regulated genome (notably including the
  methylation, acetylation etc. of the histones)
• Through the chromosomes, but outside DNA
• Gene imprinting, norms of response
• Most genetic variants have low or moderate
  penetrance (unlike the textbook cases of inborn
  errors of metabolism)
• The internal (developmental-tissutal) environment
  and the external environment may produce major
  differences, some of which are inheritable, and
  some of which are non-transitive
• Without any change in DNA sequence.

16
Enter Epigenetics

• Epigenetics the study of heritable changes in
  gene function that occur without a change in the
  DNA sequence and are therefore potentially
  reversible.

Scientific American, December 2003
17
The case of identical twins
DNA sequence identical
multiple sclerosis
asthma
18
The case of the black mouse
DNA sequence identical
The murine agouti gene encodes a signaling
molecule that signals follicular melanocytes to
switch from producing black eumelanin to yellow
phaeomelanin
19
Does nutrition matter?
Start females on diet (methyl donors and
cofactors folic acid, vit B12, anhydrous betaine)
offspring rated for coat color
mating
birth
14 d pregnancy
lactation day 21
20
A classic evolutionary puzzle

• a b good
• A B even better
• But
• a B lethal
• A b lethal
• How do you get to A B ?
• Solution You keep a and b in a DNA archive
• And then activate A and B only when both are
  available
• Silencing a and b
• An evolutionary capacitor (S. Lindquist 1998)

21
Mutation in the HSP90 genea chaperonine
22
Mutation in the HSP90 genea chaperonine, a
quality-control gene, a buffer gene
Hsp90 normally acts to reduce the likelihood
that stochastic events will alter the
deterministic unfolding of a multitude of
developmental programmes.
Rutherford, S. L., Lindquist, S. (1998). Hsp90
as a capacitor for morphological evolution.
Nature, 396 (26 November), 336-342.
23
Exposes, one generation down the line,
pre-existing silent mutations
24
Hsp90
N
N

• Ubiquitous molecular chaperone, i.e. a protein
  which binds other proteins (clients) to help
  them fold correctly.
• Essential to the activation and assembly of a
  range of client proteins typically involved in
  signal transduction, cell cycle control and
  transcriptional regulation.
• Works as a molecular clamp via transient
  dimerization of the N-terminal domains.
• Recognizes structural features common to unstable
  proteins and keeps these proteins poised for
  activation until they are stabilized by
  modifications associated with signaling.

C
ATP
ADP
C
N
N
C
C
25

• Mutation or pharmacological impairment of
  Drosophila Hsp90 leads to the emergence of
  phenotypic variation affecting nearly any adult
  structure.
• Multiple, previously silent, genetic determinants
  produce these variants and become rapidly
  independent of Hsp90 mutations.
• Widespread genetic variation affecting
  morphogenic pathways exists in nature, but is
  usually silent.
• Hsp90 buffers this variation, allowing it to
  accumulate under neutral conditions.

Nature 1998
capacitor an electric circuit element used to
store charge temporarily
26
Evo-Devo role of HSP90

• Hsp90 stabilizes its client proteins in
  conformations that would otherwise be prone to
  misfolding and potentiates their capacity to be
  activated in the proper time and place, by
  associations with partner proteins, ligand
  binding, post-translational modifications and
  correct localization.
• From the viewpoint of a population geneticist,
  buffering systems like Hsp90, without regard to
  any possible adaptive value, would decrease
  selection on nucleotide substitutions, allowing
  storage of an expanded spectrum of selectively
  nearly neutral ones. Under stable environmental
  conditions, a population arrives at a local
  fitness optimum in an adaptive landscape.
• Given that natural selection can only further
  increase the fitness of a population, it is a
  perplexing evolutionary question how a population
  might move to a different local optimum without
  an intervening period of reduced fitness
  (adaptive valley).

27
These authors punch line

• As a by-product of its biochemical function,
  Hsp90 may allow the neutral accumulation of
  potentially selectable polymorphisms and
  synchronize their conversion to a non-neutral
  state. Certainly, most combinations of
  polymorphisms will be deleterious, and these may
  be periodically purged from the population
  through the environmental coupling of the Hsp90
  buffer. However, rare combinations may produce a
  new, advantageous phenotype, thereby providing a
  molecular means by which adaptive peak shifts in
  large populations may occur without passing
  through an adaptive valley. None of the other
  mechanisms discussed genetic drift, compensatory
  mutations and gene conversion can be modulated
  by environmental contingencies. (Emphasis added)
• Christine Queitsch, Todd A. Sangster Susan
  Lindquist, Nature (2002) Vol. 417, p. 618 (the
  effects of HSP90 mutations in Arabidopsis
  thaliana)

28
A crucial lesson

• Mechanism is the name of the game!
• Without a mechanism, with descriptions only, you
  are gasping for scientific air
• The case of Waddingtons vein-less flies
• The case of Piagets lymneas
• Not to mention the un-mentionable (Lyssenko and
  his vernalization)
• What a window typically gives you Mechanisms!

29
(No Transcript)
30
The histone code
active
inactive

• Histones can be modified by chemical tags
  attached to their N-terminal tail.
• These modifications can then be interpreted by
  proteins that recognize a particular modification
  and facilitate the appropriate downstream
  biological events.

31
Epigenetic Mechanisms
methylation

• The epigenetic status changes in response to
  developmental and/or environmental cues.
• Changes rest mostly on modifications of either
  the DNA itself (methylation) or of proteins that
  intimately associate with DNA (histones? histone
  code).

acetylation methylation phosphorylation
32
Changes in chromatin accessibility dictated
bythe histone code govern gene expression
enzyme
enzyme
X
silenced
active
naïve T lymphocytes
differentiated T lymphocytes
33
Cross-talk between DNA-based andhistone-based
epigenetic mechanisms
DNA methylation Histone H3-K9
methylation
recruitment
DNA methylation Histone
deacetylation

gene silencing
34
Frequency of Epigenetic Changes

• Epigenetic explanations are quantitatively
  powerful.
• The frequency with which epigenetic modifications
  arise can reach up to 100 per locus per
  generation.
• By contrast, the rate of DNA sequence mutations
  is much lower, and varies between 105 and 109
  per locus per generation.
• A gigantic difference in frequency
• The impact of epigenetic mechanisms onto
  phylogenesis and onto a number of common diseases
  is under intense study

35
Evo-devo and the fixation of behaviors

• Evolution as the evolution of ontogenies
• The 1944-1945 famine in Holland
• Early (in ovo) auditory imprinting in birds
• Synchrony of egg-hatching and food availability
• Childhood maltreatment, monoamine-oxydase A
  alleles and aggressiveness
• 5-HTT serotonine transporter gene (LL, LS and SS
  variants) childhood experience and depression
• maternal-buffering effect in monkeys

36
An old consideration

• Due to Richard Lewontin
• New behaviors ? New environments ? New problems
  to be solved ? New selective pressures ? new
  phenotypes
• Possibly a genetic fixation of those behaviors
• The organism makes the environment
• Terrence Deacon (1997, 2003) pushes this quite
  far (E-language, society and the brain
  co-evolving)

37
A general consideration

• We have a phenotype P, such that
• Some components are universal in the species
• Many, many variants co-exist, a large but finite
  repertoire
• Specific environments canalize the choice of the
  variant
• With binary oppositions
• With critical periods (and irreversibility)
• No inheritance of the variant
• But inheritance of the universal component

38
A modern geneticists conclusion

• P cries out for an epigenetic explanation
• And an evolutionary history of fine regulations
• WELL
• Language is such a P!

39
Experience

• A new look at its nature and role

40
Inter-species conservation and parameters

• Human Otx genes have been transplanted and
  expressed in Drosophila (Leuzinger et al., 1998
  Nagao et al., 1998)
• Conversely, murine Otx genes have been replaced
  with Drosophila genes, fully rescuing
  corticogenesis impairment and epilepsy. (Acampora
  et al. 1998)
• These genes are by no means an exception.

41
Hox genes tell cells where they are along A-P axis
wildtype
Antennapedia mutant
homeotic mutations affect segment identity
42
Induction of ectopic eyes by targeted expression
of fly eyeless or its mouse homolog Pax-6.
Halder et al (1995) Science 2761788
43
eyeless master gene controlling eye development
ectopic expression of eyeless in
wing
leg
antenna
PAX6 and EY have very similar structures
44
Main lessons

• Remarkable, surprising, stunning, (quote,unquote
  from these papers by geneticists). degree of
  genetic conservation
• But also a stunning degree of individual genetic
  variation
• 150 polymorphisms per gene in humans is an
  average (including the non-coding regions) (1
  out of every 100 bases)
• Phylogenetic shadowing (SNPs) is redrawing the
  phylogenetic maps

45
The search for mental retardation genes
OMiM Online Mendelian Inheritance in Man
(database)
245
37
211
17
17
Inlow Restifo, Genetics, 2004
46
Molecular Functions of Human MR Genes
Inlow Restifo, Genetics, 2004
47
How many genes does it take to build a thinking
brain? geneticists approach How many
genes can mutate to phenotype that disrupts
cognition? eg, mental retardation Hundreds!
maybe 1,000 What biological functions do those
genes have at molecular, cellular, tissue
levels? Premise many genes required to build
the brain also required for its adult
function Almost any! How conserved are the
genetic programs across species? Remarkably well
conserved! (Slide by Prof. Linda Restifo,
Division of Neurobiology University of Arizona)
48
Parametric variation

• A murine PAX-6 gene induces a vertebrate
  (globular) eye in a vertebrate
• But a composite eye (with hundreds of ommatides,
  a rigorously fixed number) in the fruitfly,
• And vice-versa
• The parametric character of epigenetics
• The huge role of locality

49
Parameters in language
50
The logical problem of language acquisition

• Suppose the child hears one new sentence type
  every second
• How long would it take (as an average) to her to
  learn the local grammar correctly?
• Robert C. Berwick (of MIT) has made that
  calculation for a grammar with 100 rules to be
  guessed in a continuum
• 150 centuries! 15,000 years!
• This is Berwicks paradox.
• So, something totally different must be going on
• Parameter fixation as virtually conceptually
  necessary

51
A very important consideration

• The child does not have to learn that
• Parameters exist, in syntax (and/or in the
  morpho-lexicon)
• That they can each receive two possible values
  (/-)
• That each choice of the parametric value brings
  about many different syntactic effects (that
  many subtle syntactic consequences automatically
  follow).
• She/he knows all this already!
• In virtue of being human and thereby possessing
  UG.
• The only thing she/he has to learn is whether
  the local value of that parameter is or is -
• Hearing just one of those sentences is (in
  principle) enough to fix the value of that
  parameter
• In Rome, Milan, Turin etc. Fix it to
• In London, New York, Tucson etc. Fix it to -

52
Triggers, but with statistics

• See Charles Yangs and Janet Fodors work on
  language acquisition
• See Newport, Aslin, Saffran, and Gerken and
  Gomez, and Pena, Mehler, Bonatti
• Possibly dedicated statistical analyzers,
• That lubricate something like parameter setting

53
An interesting lesson

• We do have Jacobs tinkering
• But we also have quite a lot of laws of form
  (spontaneous absolute optimization)
• Genetic factors must ride on the back of these
  naturally optimal solutions (re-discovering them
  over and over)
• The whole picture of biological evolution may
  well change as a result

54
Some lessons

• High conservation of master genes suggests a lot
  of modularity
• And a lot of discrete parametric modulation of
  expression (by different tissues and by different
  environments)
• So its interesting to study modular cognitive
  precursors in distant species (singing birds,
  corvids, chinchillas, tamarins etc.)
• The DNA- sequence-centric traditional comparison
  between species is only one aspect
• Regulation, plasticity and epigenesis must be
  taken into account

55
Some challenges

• The huge degree of average individual genetic
  variation appears to clash with the very idea of
  a genetically-supported universal grammar
• If UG is genetically-supported, why is it not
  greatly variable across individuals?
• Maybe because its really minimal!
• Prying apart on a genetic basis individually
  variable components of language (the
  morpho-lexicon?) from the truly universal traits
  (NS?) is quite a challenge
• Seeing whether the very idea of discrete
  parametric language differences matches the
  epigenetic and parametric processes of brain
  maturation.

56
Back to the Minimalist Program
57
A short list of what we must have

• There are manifest expressions (there is a PHON
  component)
• There are meanings for those expressions (there
  is a SEM component)
• There is a morpho-lexicon
• There is recursion
• There is discrete infinity
• There is a combinatorics (there are syntactic
  objects composed of parts)
• There are ph(r)ases (units larger than words and
  smaller than whole sentences)
• There are internal hierarchies inside and across
  those units
• Hypothesis There is nothing else

58
The evolution of NS

• Possibly a genuinely new structure
• But possibly not much was needed
• A gene duplication, a transposon, a modified
  regulatory sequence
• Maybe with some capacitor effect
• Interfacing with improved, but basically
  pre-existing structures
• (Think of Geschwinds territory)
• Connecting them in the best possible way
• In terms of computational efficiency

59
A system driven by computational efficiency

• Signal example Copy deletion (traces)
• Why not pronounce them all?
• Witness The predicament of Brocas aphasics
• The system is driven by computational efficiency
  (maybe by anti-symmetry principles)
• Not by communicability or disambiguation
• At the opposite extreme
• Witness Structural Case, much of agreement and
  all uninterpretable features
• Utterly redundant, but maintained for some
  independent reason

60
Evidence for the hypothesis that NS is novel

• The essential mismatch between SM and CI
• Morphology is the result of a spree at one
  level of organization (sounds) (the viral
  hypothesis)
• Morphology seems to come for free
• NS is optimized for CI (not SM)
• But outward expression is (somehow) necessary
• So you have checking and feature deletion
• None of this would be the case if the system had
  evolved piecemeal, one trait conditioning the
  other and viceversa

61
Language design

• A lot makes no sense in the abstract (logically
  or generically)
• Why only have intersective quantifiers?
• Why avoid all points of symmetry?
• Why have the hierarchy of agreement shown by Mark
  Baker these days?
• Why have parametric variation? (Rather than no
  variation at all)
• Why not use parataxis (adjuncts) all the way?
• Why not have as many theta-roles as you want?
  (Hale and Keyser have an answer)
• The list can be expanded, of course

62
A lesson

• Optimal design at one level (for one component)
  may well lead to a host of elaborate consequences
  (a lot of satisficing)
• At other levels, for other components
• The selfish replication of DNA is a prominent
  example
• The immunopathologies are another
• Language seems to be yet another
• This is how biological systems work
• And evolve

63
Main innovations of the MP (SMT)

• A fundamental mismatch that has to be compensated
  for, for each and every sentence
• An excess of morphology
• And a very lean interpretive system
• Numeration as distinct from lexical insertion
• Un-valued features that have to be valued
• Not assignment any more, but checking and
  deletion
• Legibility conditions at the interfaces
• (But scant knowledge of the CI system aside from
  language)
• Overt operations more costly than covert
  operations (procrastinate)
• Endocentrity (headedness) in Merge

64
All in all

• Something like a viral infection
• From which the system has to protect itself
• And has to do it sooner rather than later
• Step by step
• Locally, with the closest candidates, with no
  look-ahead and no back-tracking
• Syntax as a limitator of expressiveness, rather
  than a facilitator of communication (Uriagereka,
  Lightfoot and yours truly)
• Jacobs tinkering and the effects of the laws of
  form seem to co-exist

65
Conclusions

• There is a sense in which Minimalism cannot be
  wrong
• Though it can be premature
• Assuming something like it does work
• Then,
• For language to evolve
• Not much had to evolve (just FLN (NS))
• And that has no obvious (or non-obvious)
  adaptationist explanation
• Its like a snow crystal
• Shaped by physics and computational efficiency
• Interacting with a lot of tinkering
• As is the case with biological systems!